Final published version
Research output: Contribution to Journal/Magazine › Journal article › peer-review
Research output: Contribution to Journal/Magazine › Journal article › peer-review
}
TY - JOUR
T1 - Breast cancer diagnosis from histopathology images using deep neural network and XGBoost
AU - Maleki, Alireza
AU - Raahemi, Mohammad
AU - Nasiri, Hamid
N1 - Publisher Copyright: © 2023 Elsevier Ltd
PY - 2023/9/1
Y1 - 2023/9/1
N2 - Background and Objective: Globally, breast cancer is one of the most common diseases among women. As a result of the disadvantages of manual analysis, computer-aided diagnosis (CAD) systems are being used to detect images because of their time-consuming and trustworthy capability. With deep learning techniques based on image analysis and classification, CAD systems can efficiently classify images. Methods: This paper proposes methodologies for enhancing the speed and precision of histopathological image classification, which is a challenge for therapeutic measures. We assess three different classifiers and six pre-trained networks. A pre-trained model is used to extract features from images and then feed those extracted features into the extreme gradient boosting (XGBoost) method, which is selected as the final classifier. Our methodology is based on transfer learning and uses histopathological images as input. To evaluate the performance of the proposed method, we use the BreakHis dataset, which presents histopathology images in four magnification levels, i.e., 40X, 100X, 200X, and 400X. Results and Conclusion: The accuracies achieved by the proposed method in 40X, 100X, 200X, and 400X magnifications are 93.6%, 91.3%, 93.8%, and 89.1%, respectively. After analyzing the accuracy achieved in this study, the final method proposed combines the DenseNet201 model as a feature extractor with XGBoost as a classifier.
AB - Background and Objective: Globally, breast cancer is one of the most common diseases among women. As a result of the disadvantages of manual analysis, computer-aided diagnosis (CAD) systems are being used to detect images because of their time-consuming and trustworthy capability. With deep learning techniques based on image analysis and classification, CAD systems can efficiently classify images. Methods: This paper proposes methodologies for enhancing the speed and precision of histopathological image classification, which is a challenge for therapeutic measures. We assess three different classifiers and six pre-trained networks. A pre-trained model is used to extract features from images and then feed those extracted features into the extreme gradient boosting (XGBoost) method, which is selected as the final classifier. Our methodology is based on transfer learning and uses histopathological images as input. To evaluate the performance of the proposed method, we use the BreakHis dataset, which presents histopathology images in four magnification levels, i.e., 40X, 100X, 200X, and 400X. Results and Conclusion: The accuracies achieved by the proposed method in 40X, 100X, 200X, and 400X magnifications are 93.6%, 91.3%, 93.8%, and 89.1%, respectively. After analyzing the accuracy achieved in this study, the final method proposed combines the DenseNet201 model as a feature extractor with XGBoost as a classifier.
KW - BreakHis
KW - Breast cancer
KW - DenseNet201
KW - Transfer learning
KW - XGBoost
U2 - 10.1016/j.bspc.2023.105152
DO - 10.1016/j.bspc.2023.105152
M3 - Journal article
AN - SCOPUS:85162128992
VL - 86
JO - Biomedical Signal Processing and Control
JF - Biomedical Signal Processing and Control
SN - 1746-8094
IS - Part A
M1 - 105152
ER -